Privacy Sandbox Topics API
Replacing Third-Party Cookies for Interest-Based Targeting
The Topics API is part of Google's Privacy Sandbox initiative and is Chrome's proposed replacement for interest-based advertising that previously relied on third-party cookies and cross-site tracking. Instead of a tracker following a user across every site via a shared cookie, the browser itself observes which sites a user visits, classifies each site into a small taxonomy of topics (e.g., "Fitness", "Travel & Transportation > Air Travel"), and periodically exposes a handful of the user's top topics to calling sites via a JavaScript API — without ever revealing the specific sites visited.
How Topic Assignment Works
Chrome maintains a fixed, publicly published taxonomy (around 470 topics in the current version) and runs an on-device classifier that maps each visited site's hostname to one or more topics locally — no browsing history ever leaves the device for this classification step. Each calling site observes at most the user's top 5 topics per week, one random topic is included with some probability specifically to limit fingerprinting precision, and topics older than a configurable number of weeks (currently a rolling 3-week window) are dropped. The browser also caps which topics a given site is even eligible to observe, based on whether that site previously called the API during the weeks those topics were being computed.
// Must be called from a top-level browsing context (or an iframe
// with the "browsing-topics" Permissions Policy explicitly allowed).
if ('browsingTopics' in document) {
const topics = await document.browsingTopics();
// Example response: array of up to 5 topic objects
// [{ topic: 143, taxonomyVersion: "2", modelVersion: "4", version: "chrome.1:2:4" }, ...]
topics.forEach(t => {
// Map t.topic (an integer ID) against the published taxonomy
// to get a human-readable category for ad selection.
});
} else {
// Fall back to contextual targeting entirely.
}Chrome Abandoned Third-Party Cookie Deprecation — Topics API Lost Its Forcing Function
Granularity Tradeoffs vs Third-Party Cookies
Topics intentionally sacrifices precision for privacy. A third-party cookie-based profile could track an individual user's specific page-level interests continuously and precisely; Topics gives you a coarse, rotating, capped set of broad categories shared with noise injected, observable only by sites the user has actually interacted with recently in a qualifying way. For advertisers used to granular retargeting ("user viewed this exact SKU"), Topics cannot replicate that — it is fundamentally an interest-category signal for broad targeting and lookalike-style audience building, not a retargeting substitute.
Where This Fits in a Post-Cookie Measurement Stack
Topics API addresses only the *targeting* side of the problem — helping an ad platform decide which ad category to show — and says nothing about *measurement* (did the ad lead to a conversion). That's handled by separate Privacy Sandbox APIs: the Attribution Reporting API for conversion measurement and the Protected Audience API (formerly FLEDGE) for remarketing-style auctions run entirely within the browser. Treat Topics, Attribution Reporting, and Protected Audience as three separate pieces of a stack, not interchangeable solutions to the same problem.
First-Party Data Still Wins Regardless of Topics API's Fate
Practical Status for MarTech Teams Today
As of now, Topics API support exists only in Chromium-based browsers (not Safari or Firefox, which have their own separate anti-tracking approaches), adoption by major DSPs and SSPs remains partial, and most production ad-targeting stacks still lean primarily on first-party data, contextual signals, and clean-room collaborations rather than Topics API output. Treat it as one input worth monitoring and testing in a sandboxed campaign, not yet a primary targeting pillar to build a strategy around.
What's Next?
The broader category of browser-level anti-tracking changes — of which Topics API is one example — has already reshaped identity resolution strategy across the industry, and that shift is worth tracing from its earlier milestones to understand why today's tools look the way they do.
I build these systems professionally.
Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.